BUILD: release FACET 1.2.1

RELEASE_NOTES.rst

@@ -10,6 +10,15 @@ fit the underlying crossfit.
 One example where this can be useful is to use only a recent period of a time series as
 the baseline of a simulation.
 
+
+1.2.1
+~~~~~
+
+- FIX: fix a bug in :class:`.UnivariateProbabilitySimulator` that was introduced in
+  FACET 1.2.0
+- catch up with FACET 1.1.1
+
+
 1.2.0
 ~~~~~
 
@@ -26,10 +35,12 @@ FACET 1.1
 FACET 1.1 refines and enhances the association/synergy/redundancy calculations provided
 by the :class:`.LearnerInspector`.
 
+
 1.1.1
 ~~~~~
 
 - DOC: add reference to FACET research paper on the project landing page
+- FIX: correctly count positive class frequency in UnivariateProbabilitySimulator
 
 
 1.1.0

src/facet/__init__.py

@@ -6,7 +6,7 @@
 """
 
 
-__version__ = "1.2.0"
+__version__ = "1.2.1"
 
 __logo__ = (
     r"""

src/facet/simulation/_simulation.py

@@ -628,11 +628,9 @@ def expected_output(self) -> float:
 
         :return: observed frequency of the positive class
         """
-        actual_outputs = self.sample.target.loc[self.subsample]
+        actual_outputs = self.sample.target
 
-        return actual_outputs.loc[actual_outputs == self._positive_class()].sum() / len(
-            actual_outputs
-        )
+        return (actual_outputs == self._positive_class()).sum() / len(actual_outputs)
 
     def _positive_class(self) -> Any:
         """

test/test/conftest.py

@@ -11,7 +11,8 @@
 from sklearn.utils import Bunch
 
 from sklearndf import TransformerDF
-from sklearndf.pipeline import RegressorPipelineDF
+from sklearndf.classification import RandomForestClassifierDF
+from sklearndf.pipeline import ClassifierPipelineDF, RegressorPipelineDF
 from sklearndf.regression import (
     SVRDF,
     AdaBoostRegressorDF,
@@ -349,3 +350,86 @@ def check_ranking(
                 f"unexpected parameters for learner at rank #{rank}: "
                 f"got {parameters_actual} but expected {parameters_expected}"
             )
+
+
+@pytest.fixture
+def iris_classifier_ranker_binary(
+    iris_sample_binary: Sample,
+    cv_stratified_bootstrap: StratifiedBootstrapCV,
+    n_jobs: int,
+) -> LearnerRanker[ClassifierPipelineDF[RandomForestClassifierDF]]:
+    return fit_learner_ranker(
+        sample=iris_sample_binary, cv=cv_stratified_bootstrap, n_jobs=n_jobs
+    )
+
+
+@pytest.fixture
+def iris_classifier_ranker_multi_class(
+    iris_sample: Sample, cv_stratified_bootstrap: StratifiedBootstrapCV, n_jobs: int
+) -> LearnerRanker[ClassifierPipelineDF[RandomForestClassifierDF]]:
+    return fit_learner_ranker(
+        sample=iris_sample, cv=cv_stratified_bootstrap, n_jobs=n_jobs
+    )
+
+
+@pytest.fixture
+def iris_classifier_ranker_dual_target(
+    iris_sample_binary_dual_target: Sample, cv_bootstrap: BootstrapCV, n_jobs: int
+) -> LearnerRanker[ClassifierPipelineDF[RandomForestClassifierDF]]:
+    return fit_learner_ranker(
+        sample=iris_sample_binary_dual_target, cv=cv_bootstrap, n_jobs=n_jobs
+    )
+
+
+@pytest.fixture
+def iris_classifier_crossfit_binary(
+    iris_classifier_ranker_binary: LearnerRanker[ClassifierPipelineDF],
+) -> LearnerCrossfit[ClassifierPipelineDF[RandomForestClassifierDF]]:
+    return iris_classifier_ranker_binary.best_model_crossfit_
+
+
+@pytest.fixture
+def iris_classifier_crossfit_multi_class(
+    iris_classifier_ranker_multi_class: LearnerRanker[ClassifierPipelineDF],
+) -> LearnerCrossfit[ClassifierPipelineDF[RandomForestClassifierDF]]:
+    return iris_classifier_ranker_multi_class.best_model_crossfit_
+
+
+@pytest.fixture
+def iris_inspector_multi_class(
+    iris_classifier_crossfit_multi_class: LearnerCrossfit[
+        ClassifierPipelineDF[RandomForestClassifierDF]
+    ],
+    n_jobs: int,
+) -> LearnerInspector[ClassifierPipelineDF[RandomForestClassifierDF]]:
+    return LearnerInspector(shap_interaction=True, n_jobs=n_jobs).fit(
+        crossfit=iris_classifier_crossfit_multi_class
+    )
+
+
+#
+# Utility functions
+#
+
+
+def fit_learner_ranker(
+    sample: Sample, cv: BaseCrossValidator, n_jobs: int
+) -> LearnerRanker[ClassifierPipelineDF[RandomForestClassifierDF]]:
+    # define parameters and crossfit
+    grids = [
+        LearnerGrid(
+            pipeline=ClassifierPipelineDF(
+                classifier=RandomForestClassifierDF(random_state=42), preprocessing=None
+            ),
+            learner_parameters={"n_estimators": [10, 50], "min_samples_leaf": [4, 8]},
+        )
+    ]
+    # pipeline inspector does only support binary classification - hence
+    # filter the test_sample down to only 2 target classes:
+    return LearnerRanker(
+        grids=grids,
+        cv=cv,
+        scoring="f1_macro",
+        random_state=42,
+        n_jobs=n_jobs,
+    ).fit(sample=sample)

test/test/facet/test_inspection.py

@@ -10,7 +10,7 @@
 import pytest
 from pandas.testing import assert_frame_equal, assert_series_equal
 from sklearn.datasets import make_classification
-from sklearn.model_selection import BaseCrossValidator, KFold
+from sklearn.model_selection import KFold
 
 from pytools.viz.dendrogram import DendrogramDrawer, DendrogramReportStyle
 from sklearndf import TransformerDF
@@ -29,7 +29,7 @@
     TreeExplainerFactory,
 )
 from facet.selection import LearnerGrid, LearnerRanker
-from facet.validation import BootstrapCV, StratifiedBootstrapCV
+from facet.validation import BootstrapCV
 
 # noinspection PyMissingOrEmptyDocstring
 
@@ -38,61 +38,6 @@
 T = TypeVar("T")
 
 
-@pytest.fixture
-def iris_classifier_ranker_binary(
-    iris_sample_binary: Sample,
-    cv_stratified_bootstrap: StratifiedBootstrapCV,
-    n_jobs: int,
-) -> LearnerRanker[ClassifierPipelineDF[RandomForestClassifierDF]]:
-    return fit_learner_ranker(
-        sample=iris_sample_binary, cv=cv_stratified_bootstrap, n_jobs=n_jobs
-    )
-
-
-@pytest.fixture
-def iris_classifier_ranker_multi_class(
-    iris_sample: Sample, cv_stratified_bootstrap: StratifiedBootstrapCV, n_jobs: int
-) -> LearnerRanker[ClassifierPipelineDF[RandomForestClassifierDF]]:
-    return fit_learner_ranker(
-        sample=iris_sample, cv=cv_stratified_bootstrap, n_jobs=n_jobs
-    )
-
-
-@pytest.fixture
-def iris_classifier_ranker_dual_target(
-    iris_sample_binary_dual_target: Sample, cv_bootstrap: BootstrapCV, n_jobs: int
-) -> LearnerRanker[ClassifierPipelineDF[RandomForestClassifierDF]]:
-    return fit_learner_ranker(
-        sample=iris_sample_binary_dual_target, cv=cv_bootstrap, n_jobs=n_jobs
-    )
-
-
-@pytest.fixture
-def iris_classifier_crossfit_binary(
-    iris_classifier_ranker_binary: LearnerRanker[ClassifierPipelineDF],
-) -> LearnerCrossfit[ClassifierPipelineDF[RandomForestClassifierDF]]:
-    return iris_classifier_ranker_binary.best_model_crossfit_
-
-
-@pytest.fixture
-def iris_classifier_crossfit_multi_class(
-    iris_classifier_ranker_multi_class: LearnerRanker[ClassifierPipelineDF],
-) -> LearnerCrossfit[ClassifierPipelineDF[RandomForestClassifierDF]]:
-    return iris_classifier_ranker_multi_class.best_model_crossfit_
-
-
-@pytest.fixture
-def iris_inspector_multi_class(
-    iris_classifier_crossfit_multi_class: LearnerCrossfit[
-        ClassifierPipelineDF[RandomForestClassifierDF]
-    ],
-    n_jobs: int,
-) -> LearnerInspector[ClassifierPipelineDF[RandomForestClassifierDF]]:
-    return LearnerInspector(shap_interaction=True, n_jobs=n_jobs).fit(
-        crossfit=iris_classifier_crossfit_multi_class
-    )
-
-
 def test_model_inspection(
     regressor_grids: Sequence[LearnerGrid[RegressorPipelineDF]],
     regressor_ranker: LearnerRanker[RegressorPipelineDF],
@@ -815,29 +760,6 @@ def test_shap_plot_data(
 #
 
 
-def fit_learner_ranker(
-    sample: Sample, cv: BaseCrossValidator, n_jobs: int
-) -> LearnerRanker[ClassifierPipelineDF[RandomForestClassifierDF]]:
-    # define parameters and crossfit
-    grids = [
-        LearnerGrid(
-            pipeline=ClassifierPipelineDF(
-                classifier=RandomForestClassifierDF(random_state=42), preprocessing=None
-            ),
-            learner_parameters={"n_estimators": [10, 50], "min_samples_leaf": [4, 8]},
-        )
-    ]
-    # pipeline inspector does only support binary classification - hence
-    # filter the test_sample down to only 2 target classes:
-    return LearnerRanker(
-        grids=grids,
-        cv=cv,
-        scoring="f1_macro",
-        random_state=42,
-        n_jobs=n_jobs,
-    ).fit(sample=sample)
-
-
 def print_expected_matrix(error: AssertionError, split: bool = False):
     # used to print expected output for copy/paste into assertion statement
 

test/test/facet/test_simulation.py

@@ -7,13 +7,15 @@
 from pytest import approx
 
 from sklearndf import TransformerDF
-from sklearndf.pipeline import RegressorPipelineDF
+from sklearndf.classification import RandomForestClassifierDF
+from sklearndf.pipeline import ClassifierPipelineDF, RegressorPipelineDF
 from sklearndf.regression.extra import LGBMRegressorDF
 
 from facet.crossfit import LearnerCrossfit
 from facet.data import Sample
 from facet.data.partition import ContinuousRangePartitioner
 from facet.simulation import (
+    UnivariateProbabilitySimulator,
     UnivariateSimulationResult,
     UnivariateTargetSimulator,
     UnivariateUpliftSimulator,
@@ -477,8 +479,62 @@ def test_univariate_uplift_subsample_simulation(
         simulation_result.partitioner.frequencies_, [1, 4, 9, 10, 10, 6, 2, 1, 4]
     )
 
-    SimulationDrawer(style="text").draw(
-        data=uplift_simulator.simulate_feature(
-            feature_name=parameterized_feature, partitioner=partitioner
-        )
+    SimulationDrawer(style="text").draw(data=simulation_result)
+
+
+def test_univariate_probability_simulation(
+    iris_classifier_crossfit_binary: LearnerCrossfit[
+        ClassifierPipelineDF[RandomForestClassifierDF]
+    ],
+    n_jobs: int,
+) -> None:
+    parameterized_feature = "sepal length (cm)"
+    partitioner = ContinuousRangePartitioner(max_partitions=10)
+
+    print(iris_classifier_crossfit_binary.sample_.feature_names)
+
+    proba_simulator = UnivariateProbabilitySimulator(
+        crossfit=iris_classifier_crossfit_binary,
+        confidence_level=0.95,
+        n_jobs=n_jobs,
+        verbose=50,
+    )
+
+    simulation_result: UnivariateSimulationResult = proba_simulator.simulate_feature(
+        feature_name=parameterized_feature, partitioner=partitioner
     )
+
+    index = pd.Index(
+        data=[5, 5.5, 6, 6.5, 7, 7.5, 8], name=UnivariateUpliftSimulator.IDX_PARTITION
+    )
+
+    assert simulation_result.baseline == approx(0.5)
+
+    assert_series_equal(
+        simulation_result.outputs_lower_bound(),
+        pd.Series(
+            [0.346255, 0.346255, 0.353697, 0.394167, 0.401895, 0.417372, 0.417372],
+            name=UnivariateSimulationResult.COL_LOWER_BOUND,
+            index=index,
+        ),
+    )
+
+    assert_series_equal(
+        simulation_result.outputs_median(),
+        pd.Series(
+            [0.460432, 0.450516, 0.469412, 0.488569, 0.492651, 0.507788, 0.507788],
+            name=UnivariateSimulationResult.COL_MEDIAN,
+            index=index,
+        ),
+    )
+
+    assert_series_equal(
+        simulation_result.outputs_upper_bound(),
+        pd.Series(
+            [0.582565, 0.562096, 0.570590, 0.580023, 0.599714, 0.602303, 0.602303],
+            name=UnivariateSimulationResult.COL_UPPER_BOUND,
+            index=index,
+        ),
+    )
+
+    SimulationDrawer(style="text").draw(data=simulation_result)